Biometric Policy

Sidekick Live
ClassDojo offers an optional on-device classroom audio recording and transcription feature (“Sidekick Live”) that helps teachers capture and analyze classroom discussions for educational purposes. When a teacher opts in and initiates a recording, this feature records classroom audio (including student voices) on the teacher’s device and converts it to a text transcript using Apple’s on-device speech recognition engine implemented using Apple’s SpeechAnalyzer / SpeechTranscriber APIs (“Classroom Audio Transcript”). Raw audio is processed in memory on the teacher’s device during the active session and is not written to an audio file. ClassDojo does not create, collect, or store a raw audio recording file. When the session ends, no saved audio file remains on the teacher’s device, so the teacher does not need to delete one. The transcription of the audio also takes place on the teacher’s device using Apple’s on-device speech recognition (using the SpeechAnalyzer APIs) and may involve the use of AI to create that transcription. After the audio has been transcribed on-device, ClassDojo collects only the resulting Classroom Audio Transcript (text only) from the teacher’s device and sends the transcript through ClassDojo’s AI Classroom Tools for further processing, including processing by ClassDojo’s Large Language Model (LLM) service providers (“LLM Service Providers”), to create AI outputs (such as summaries of the transcripts). For a list of current LLM Service Providers, see ClassDojo’s Third-Party Service Provider page.

Teachers must opt in to activate Sidekick Live and must confirm that: (1) their school or district allows such recordings; and (2) they have obtained all necessary consents in connection with the recording, including without limitation, any consents required to satisfy applicable two-party or all-party consent recording laws and any consents required for the collection of biometric data such as voice recordings.

ClassDojo does not collect or store the raw audio or voice recording. ClassDojo also does not collect or store any of the on-device processing done by Apple to turn the audio into a transcript. After the audio has been transcribed, ClassDojo then collects the output produced by Apple on the device (the text Classroom Audio Transcript) and sends the transcript to its LLM Service Providers for additional processing to create AI outputs (such as summaries of the transcripts). For more information on the “input” from teachers, School Leaders, and Admins, and “prompts” (or instructions) submitted by ClassDojo and used to generate the AI “outputs”, as well as the information shared with our LLM Service Providers in connection with Sidekick Live, please see our AI Transparency Note. 

Live Points
ClassDojo also offers a separate optional live classroom audio feature (“Live Points”) designed to let a teacher award points or trigger classroom behavior acknowledgments through spoken commands while class is in session. While Live Points is active, the ClassDojo mobile app captures live microphone audio on the teacher’s device, uses Apple’s on-device speech-recognition (SFSpeechRecognizer) tools to generate recognized words or phrases entirely on the teacher’s device, and then the ClassDojo app locally compares those recognized words or phrases against predefined command patterns. The recognition results can be visible to the ClassDojo app locally on the teacher’s device without being separately sent to ClassDojo’s servers.

Teachers must opt in to activate Live Points and must confirm that: (1) their school or district allows such recordings; and (2) they have obtained all necessary consents in connection with the recording, including without limitation, any consents required to satisfy applicable two-party or all-party consent recording laws and any consents required for the collection of biometric data such as voice recordings.

ClassDojo has configured Apple's on device speech recognition so that audio stays on the teacher's device and is not sent to Apple over the internet. Additionally, the audio is not sent to or saved by ClassDojo. Apple states that this setting keeps audio data on the device, and Apple’s WWDC23 materials state that customized requests are serviced strictly on device and customization data is never sent over the network. Accordingly, Apple does not receive the audio, transcription output, or customization data through this speech-recognition flow and cannot use that data for model training. Additionally, Apple’s SFSpeechRecognizer API distinguishes between captured audio content from the device’s microphone and a recorded audio file. Live Points is configured to use the live-audio path, such that the microphone audio is processed as transient data held temporarily on the teacher’s device during recognition rather than being written to a persistent audio file on the teacher’s device. Transient local retention of audio data may occur during processing while Live Points is active, and additional live audio capture for this feature stops when either ClassDojo or the teacher stops the live-audio session.

ClassDojo has also configured a custom vocabulary of classroom terms — such as student names and behavior labels — to help Apple's speech recognition better understand what the teacher is saying. Apple states that this customization process also stays entirely on the teacher's device: "LM customization supports this by never sending customization data over the network. All customized requests are serviced strictly on device". Because the audio, the recognized words, and the custom vocabulary all remain on the teacher's device, Apple does not receive any of this data and cannot use it for any purpose, including training Apple's own speech models. Any local custom language model or vocabulary files used for this feature are separate from the transient raw audio described above and are created and used locally on the teacher’s device.

Unlike Sidekick Live, Live Points is not designed to create a classroom-discussion transcript for later AI summarization and does not send an ongoing classroom transcript through ClassDojo’s LLM Service Providers. Instead, Live Points uses local rule-based transcript matching (“pattern matching”). In plain language, this means the software compares locally generated words or partial phrases against expected terms and command patterns — such as student names from the class roster, pre-set behaviors/skills, aliases, pronunciation aids, command templates, and related structured vocabulary supplied by ClassDojo and/or the teacher or school — to determine whether a valid point-award command was spoken. ClassDojo has configured the request to require on-device recognition. If no qualifying match is detected, no transcript is sent to ClassDojo’s servers, and the locally generated recognized words or phrases used for matching are intended to be discarded after use.

While Live Points is active, Apple’s on-device speech-recognition continuously listens to live classroom audio from the teacher’s device microphone and converts it into recognized words and phrases on the teacher’s device so the Classroom app can check for matching commands. This means that while the feature is active, all audible speech in the classroom — not only the teacher's voice — is continuously converted to text by Apple's on-device speech recognition on the teacher's device for local command matching. The ClassDojo app then compares the recognized words and phrases against predefined command patterns (such as a student name paired with a behavior term) to determine whether a valid point-award command was spoken. This processing occurs entirely on the teacher's device. No classroom audio and no recognized words or phrases are sent to Apple's servers, ClassDojo's servers, or any third party. Unlike cloud-based voice assistants such as Alexa or Google Assistant, Live Points does not stream audio to any remote server for processing. However, Live Points is not a wake-word detector that listens only for a single trigger phrase and ignores everything else. If no matching command is recognized, the recognized words are discarded on the teacher's device and no data leaves the device. If a matching command is recognized, only the structured point-award data (the same information transmitted during a manual point award) is sent to ClassDojo so the point can be awarded — not the audio, not a transcript, and not the recognized words that led to the match. Teachers can stop this processing at any time by turning off Live Points.

Sidekick Live

Sidekick Live is offered only where permitted by applicable law and where the relevant school or district has authorized the feature. The Apple SpeechAnalyzer/SpeechTranscriber also depends on a supported Apple device, including a supported transcription language, microphone permission, speech-recognition authorization, and availability of the relevant Apple speech model asset of the device. Sidekick Live requires iOS 26 or later, if the teacher’s device does not support this iOS version, Sidekick Live will not be available on that device. If school or district policy, applicable state wiretapping law, applicable biometric data law, or other applicable law restricts classroom recording or voice data processing, Sidekick Live will not be available. Teachers must opt-in to starting the individual session each time they activate the feature.

Live Points

Live Points is offered only where permitted by applicable law and where the relevant school or district has authorized the feature. The feature requires a supported Apple device, microphone permission, speech-recognition authorization, and on-device speech recognition support and availability for the relevant device and language. Live Points also depends on access to the applicable class roster and behavior/skill vocabulary needed for local matching accuracy and, if used by the implementation, to build the local custom language model. Any custom language model and vocabulary files used for this feature are created and used locally on the teacher’s device through Apple’s on-device customization path; they are not downloaded from Apple. If on-device speech recognition is not available for the relevant device or language, Live Points will not function until on-device recognition becomes available. Live Points does not fall back to server-based speech recognition.

Sidekick Live

On-Device Processing and Apple Speech Framework
When a teacher activates Sidekick Live, the teacher’s device captures classroom audio through its microphone and processes that audio in memory on the device through Apple’s Speech framework (SpeechAnalyzer / SpeechTranscriber). ClassDojo does not create, collect or store a raw audio or voice recording. When the session ends, no saved audio file remains on the teacher’s device, so the teacher does not need to delete one. The transcription of the audio (the Classroom Audio Transcript) also takes place on the teacher’s device through Apple’s Speech framework and may involve the use of AI to create that transcription. Apple states that its on-device speech recognition processes audio locally on the device’s hardware without transmitting audio data to Apple or any third party. Apple does not receive, access, or store the audio or resulting transcript. Apple also states that the relevant model assets may be downloaded as needed, retained in system storage on the device, and updated automatically by the system.

Classroom Audio Transcript and Zero Data Retention Processing
ClassDojo also does not collect or store any of the intermediate on-device processing done by Apple to turn the audio into a transcript. Only the resulting text Classroom Audio Transcript (not the audio) is transmitted to ClassDojo. ClassDojo then processes the Classroom Audio Transcript through its third-party LLM Service Providers under a Zero Data Retention (ZDR) agreement to create AI outputs (such as summaries of the transcripts). Under the ZDR agreement: (1) prompts, inputs, and outputs will not be logged for human review by the LLM Service Providers; and (2) prompts, inputs, and outputs will not be retained by the LLM Service Provider beyond the duration of the active API call. The LLM Service Provider may perform automated screening of inputs and prompts for safety purposes (“Safety Classifiers”). Safety Classifiers consist solely of metadata (including classifier types, dates, counts, and confidence scores) and do not include any inputs, prompts, or outputs (or any portion or summarization thereof). LLM Service Providers may retain Safety Classifier metadata.

For more information on the “input” from teachers, School Leaders, and Admins, and “prompts” (or instructions) submitted by ClassDojo and used to generate the AI “outputs”, as well as the information shared with our LLM Service Providers in connection with Sidekick Live, please see our AI Transparency Note. 

The use of classroom audio recording and transcription also raises separate legal and policy requirements, including biometric privacy, recording-consent, student-privacy, and school-authorization issues. See the “Biometric Privacy, Recording, COPPA, FERPA, and District Requirements” section below.

Live Points

On-Device Speech Processing, Local Vocabulary, and Pattern Matching
When a teacher activates Live Points, the teacher’s device captures live classroom audio through its microphone and uses Apple’s SFSpeechRecognizer API, together with a custom speech language model, to convert speech into local text or partial text on the device. The feature requires on-device recognition so the audio stays on the device during recognition.

To improve accuracy for classroom-specific language, the app may load or generate local customization inputs using student roster names, behavior/skill labels, aliases, pronunciation data, and command templates. These inputs are used to build or prepare a local custom speech language model and related vocabulary used by the speech recognizer to better understand the names and behavior terms that are specific to that teacher’s class. Apple states that custom language model requests are serviced strictly on device and that customization data is never sent over the network, although relevant compiled language-model or vocabulary files may exist locally on the device.

The feature then applies local pattern matching to the locally generated text. “Pattern matching” means the software compares the locally generated words or phrases against expected command structures — for example, combinations of a student name plus a behavior/skill term — and may apply simple rules and confidence thresholds to determine whether a valid point-award command was likely spoken. This is not speaker recognition and does not use a generative LLM, but may use AI.

If the system cannot confidently identify a student or a behavior/skill, the current product behavior may be a silent failure or internal “no match” result rather than a point award.

Live Audio Path and Transient Processing
Apple’s SFSpeechRecognizer API distinguishes between live audio captured from the device’s microphone and a prerecorded audio file. ClassDojo has configured Live Points to use the live-audio path, such that microphone audio is processed as transient data held temporarily on the teacher’s device rather than being saved as a persistent audio recording file on the teacher’s device. This temporary retention occurs only while Live Points is active. When the teacher stops or pauses Live Points, the app stops delivering additional live audio for the session by removing the microphone buffer handler, signaling end-of-stream to Apple’s live speech-recognition request, canceling the active speech-recognition task, and discarding the per-session voice service. No additional live audio is processed for that session after Live Points is stopped or paused.

Scope of Audio Processing
While Live Points is active, the ClassDojo mobile app uses Apple's on-device speech recognition to continuously listen to live classroom audio from the teacher's device microphone and convert it into recognized words and phrases - all on the teacher's device. This means that while the feature is active, all audible speech in the classroom — not only the teacher's voice — is continuously converted to text by Apple's on-device speech recognition on the teacher's device so that the ClassDojo app can check it for matching commands. Live Points is not a wake-word detector that listens only for a single trigger phrase and ignores everything else. While the feature is on, all audible classroom speech is continuously converted to text on the teacher’s device so that the app can check it for matching commands. No classroom audio is sent to Apple's servers, ClassDojo's servers, or any third party. The recognized words or phrases are used locally on the teacher’s device for matching and are not sent off the device. Unlike cloud-based voice assistants such as Alexa or Google Assistant, Live Points does not stream audio to any remote server for processing. Teachers can stop this processing at any time by turning off Live Points.

Match and No-Match Behavior
If no valid command is recognized, the recognized words are discarded on the teacher’s device and no data leaves the device. If a valid command is recognized, the app sends only the structured point-award data fields to ClassDojo that would be sent if the teacher had manually tapped to award the point — such as the relevant student or group identifier, point value, and behavior category. No audio, transcript text, phrase snippets, confidence scores, or speech metadata are sent to ClassDojo servers as part of that point-award request. The recognized text is used purely client-side for intent parsing and is discarded after the action is dispatched. Unlike Sidekick Live, Live Points does not create a classroom-discussion transcript for later AI summarization and does not send an ongoing classroom transcript through ClassDojo’s LLM Service Providers. No generative large language model (LLM) is involved in the matching or auto-award step and the matching does not involve the use of ClassDojo’s LLM Service Providers.

Sidekick Live

1. Classroom audio, captured on the teacher’s device microphone only. The audio recording contains vocal characteristics of every person speaking in the classroom, including the teacher and students. The audio is processed in memory on-device through Apple’s on-device speech recognition engine implemented using Apple’s SpeechAnalyzer / SpeechTranscriber APIs and is not written to an audio file. ClassDojo does not separately create, collect or store the audio for Sidekick Live.

2. Classroom Audio Transcript of the classroom discussion. The Classroom Audio Transcript (text only) may incidentally contain student names, statements, and other personal information spoken during class. The Classroom Audio Transcript is transmitted to ClassDojo and then further transmitted by ClassDojo to our LLM Service Providers to provide the output (e.g. summaries of the Classroom Audio Transcripts) which are provided back to the teacher.

3. AI-generated analysis (or “output”) of the Classroom Audio Transcript, returned by the LLM Service Providers to ClassDojo (such as summaries of the transcripts). For more information on the “input” from teachers, School Leaders, and Admins, and “prompts” (or instructions) submitted by ClassDojo and used to generate the AI “outputs”, as well as the information shared with our LLM Service Providers in connection with Sidekick Live, please see our AI Transparency Note. 
   
   Depending on the jurisdiction, the audio recording and transcription process may involve the collection or processing of biometric identifiers and biometric information under applicable state laws, personal information under Children’s Online Privacy Protection Act (COPPA), sensitive personal data under applicable state privacy laws, and education records under the Family Educational Rights and Privacy Act (FERPA) when associated with a student. Audio recording is also subject to applicable state and federal wiretapping and electronic surveillance laws as described below. See the “Biometric Privacy, Recording, COPPA, FERPA and District Policies” Section below for more information.

Live Points

1. Live classroom audio temporarily captured on the teacher’s device while Live Points is active. While the feature is active, all audible speech near the teacher’s device microphone is processed on-device, including the voices of students and other persons present in the classroom, not only the teacher’s voice. No audio is written to disk, the audio is captured as in-memory buffers and processed entirely in memory.

2. Ephemeral local text and/or partial transcription generated on device solely to determine whether a valid point-award command was spoken. Unlike Sidekick Live, this local matching text is not intended to be transmitted as an ongoing transcript for LLM analysis.

3. Local customization inputs used to improve recognition accuracy, including student roster names, behavior/skill lists, aliases, pronunciations, command templates, and other structured vocabulary or language-model data associated with the relevant class, as well as related temporary on-device vocabulary or speech-model files generated from that information.

4. Matching results / point-award event data. If a qualifying command is detected, the ClassDojo app generates the standard structured point-award event fields needed to record the point (for example the relevant student or group identifier, point value, and behavior/skill category) and synchronizes that result with ClassDojo so the point can be recorded. The recognized words or phrases used to identify the command can be visible locally to the ClassDojo app on the teacher’s device for matching, but no transcript text, phrase snippets, confidence scores, or speech metadata are sent to ClassDojo servers as part of that point-award event. The recognized text is used purely client-side for intent parsing and is discarded after the action is dispatched.
   
   Depending on the jurisdiction, the live audio capture and on-device processing described above may involve the collection or processing of biometric identifiers and biometric information under applicable state laws, personal information under COPPA, and education records under FERPA when associated with a student. Live audio capture may also be subject to applicable wiretapping and electronic surveillance laws. See the “Biometric Privacy, Recording, COPPA, FERPA and District Policies” section below for more information.

Sidekick Live
Solely to enable the Sidekick Live feature as requested by teachers —to provide teachers with summaries, observations, and analysis of classroom discussions for educational purposes. We will not use this data for any other purpose.

Live Points
Solely to enable hands-free or voice-assisted awarding of classroom points, reduce classroom disruption, and improve recognition accuracy for classroom-specific names and behavior terms. Live Points is not intended to summarize classroom discussions or to create general-purpose generative AI outputs.

Sidekick Live and Live Points
AI Training Commitment
We will not use the audio, Classroom Audio Transcript, Live Points audio, or any biometric data captured during this process to train any of ClassDojo’s artificial intelligence (AI) systems.

ClassDojo does not use any children’s or teen’s personal information contained in any inputs, prompts, or output to train any of ClassDojo’s AI systems. ClassDojo will only train our AI systems with data that has been de-identified  in such a way that is not traceable to any ClassDojo user. Additionally, ClassDojo does not allow any LLM Service Provider or AI third-party service provider it may utilize to use any information (personal information or de-identified) to train, improve or develop their AI systems. Live Points does not use ClassDojo’s LLM Service Providers to analyze ongoing classroom audio or local matching text. This commitment includes the information contained in the chart detailed in the AI Transparency Note (in school use) and Parent and Child AI Transparency Page (out of school use), such as the teacher, parent, child or teen provided inputs, used to produce the AI outputs. For more information about our retention periods for this information, please see this chart for in school use and this chart for out of school use.

Sidekick Live

Classroom audio: The raw audio recording is processed in memory on the teacher’s device during the active session and is not received or stored by ClassDojo as an audio file. No raw audio recording file is created by ClassDojo for Sidekick Live and no audio recording is stored by ClassDojo, Apple, or any third party at any point. When the session ends, the app stops delivering additional audio. Teachers do not need to delete a saved audio file because none is created. Apple’s published SpeechAnalyzer materials state that the relevant transcription model assets may be downloaded to and retained in system storage on the device.

Classroom Audio Transcript: ClassDojo retains the text transcript only transiently for transmission to, and processing by, its LLM Service Providers. The LLM Service Provider retains the transcript only transiently for processing—it is immediately discarded after the analysis response is returned and subject to a ZDR agreement mentioned above.

Consent and compliance records: Consent records, withdrawal logs, and suppression records are retained as needed to document authorization choices and honor requests.

Live Points

Audio: Live audio is processed locally on the teacher’s device during active recognition and is not received or stored by ClassDojo as a raw audio file. When Live Points is stopped or paused, the app stops delivering additional audio to the live recognition request and ends the active speech-recognition task.

Ephemeral local text / partial transcription: Used locally for match evaluation and discarded after use. If a valid command is recognized, no transcript text, phrase snippets, confidence scores, or speech metadata are sent to ClassDojo servers as part of the point-award request.

Local customization inputs / model files: Student roster names, behavior/skill terms, pronunciations and templates may be used to create temporary on-device vocabulary and speech-model files. Those files, including compiled language-model or vocabulary files, may be retained, reused, or cached locally on the device during the session to help the device recognize classroom-specific names and commands more accurately. These files may contain roster names, behavior terms, pronunciations, templates, and related language-model data — not audio — and are deleted when the Live Points voice command service is torn down at the end of the session.

Point-award records: If a qualifying command results in a point award, the retained record is the same structured point-award event data that would be generated for a manual point award, subject to the applicable school agreement and ordinary retention rules for point-award records. See our FAQ for more information on retention.

Sidekick Live

We do not sell, lease, or trade any voice data, audio recordings, or Classroom Audio Transcripts. ClassDojo does not report, transmit, or include any Classroom Audio Transcripts, voice data, or biometric data in any state or federal longitudinal student data system. The audio is never shared with anyone—it remains on the teacher’s device, is processed in memory during the active session, and is not written to an audio file. The Classroom Audio Transcript (text) is shared with ClassDojo and then ClassDojo only shares the Classroom Audio Transcript with our LLM Service Providers which process the Classroom Audio Transcript solely to return the analysis to ClassDojo, under a written ZDR agreement which provides that: (1) prompts, inputs, and outputs will not be logged for human review by the LLM Service Providers; and (2) prompts, inputs, and outputs will not be retained by the LLM Service Provider beyond the duration of the active API call. The LLM Service Provider may perform automated screening of inputs and prompts for safety purposes (“Safety Classifiers”). Safety Classifiers consist solely of metadata (including classifier types, dates, counts, and confidence scores) and do not include any inputs, prompts, or outputs (or any portion or summarization thereof). LLM Service Providers may retain Safety Classifier metadata.

Live Points

We do not sell, lease, or trade any Live Points voice data, audio recordings, or locally generated matching text. ClassDojo does not report, transmit, or include any Live Points voice data, locally generated matching text, or biometric data in any state or federal longitudinal student data system. Live Points raw audio and local matching text are not shared with ClassDojo’s LLM Service Providers and are not part of the Sidekick Live transcript-to-LLM workflow. The live audio remains on the teacher’s device during local speech recognition and pattern matching. Any custom language model and vocabulary files used for Live Points are created and used locally on the teacher’s device through Apple’s on-device customization path. If a valid point-award match is detected, ClassDojo receives only the same limited structured event data that would be sent during a manual point award by a teacher, such as the applicable student or group identifier, point value, and behavior category, solely so the point can be recorded. No audio, transcript text, phrase snippets, confidence scores, or speech metadata are sent to ClassDojo servers as part of that point-award request. The recognized text is used solely on the teacher’s device for local intent parsing and is discarded after the action is dispatched.

Sidekick Live

Sidekick Live is entirely optional. The feature is off by default and requires teacher authorization each session. The teacher must affirmatively activate the audio recording via an in-app consent flow and may stop the recording at any time by tapping the stop button. The teacher’s school or district can separately disable Sidekick Live feature. In addition, the teacher may stop the transcription of the audio into the Classroom Audio Transcript. Classroom Audio Transcripts are not stored by ClassDojo and are retained only transiently so they can be transmitted to, and processed, by ClassDojo’s LLM Service Providers.

School Admins can request deletion of the output produced from the use of the ClassDojo AI Classroom Tools with Sidekick Live (e.g. the summaries of the Classroom Audio Transcripts) at any time through the app or by contacting us at privacy@classdojo.com. **Note that the deletion by a school admin of any output produced from the use of Sidekick Live may not be fully deleted from the ClassDojo servers until we receive direction from the district representative to delete pursuant to the district’s applicable Student Data Protection Addendum. Declining to use this feature does not affect the teacher’s ability to use any other ClassDojo feature.

Live Points

Live Points is entirely optional. The feature is off by default, and available only after the teacher initiates the feature for each class. The teacher must affirmatively activate the feature before the device begins listening for Live Points commands and is able to stop the feature at any time. Teachers can also delete any points automatically awarded through Live Points. Because Live Points relies primarily on local processing, the principal retained data, if any, are generally the resulting point-award records and any related logs. Declining to use this feature does not affect the teacher’s ability to use any other ClassDojo feature.

The teacher’s school or district can separately disable the Live Points feature. To request deletion of any Live Points-related records or data, schools and districts may contact ClassDojo at privacy@classdojo.com. Because Live Points processes data locally on the teacher’s device, the physical security of the device is important. Any locally stored vocabulary or language model files used for Live Points recognition accuracy are subject to the device’s own security protections (such as passcode, Face ID, or Touch ID). Teachers should ensure their devices are secured with a passcode or biometric lock.

Sidekick Live and Live Points

Depending on your state of residence, you may have additional rights regarding voice data, biometric data, student personal information, retained Live Point records, and voice recordings, including the right to access, correct, delete, and restrict processing. See our full Privacy Policy for complete details.

Note on Live Points

Live Points, like Sidekick Live, involves the capture and on-device processing of live classroom voice audio, including the voices of students present in the classroom. Although Live Points differs from Sidekick Live in significant operational ways — it uses Apple’s on-device speech-recognition tools (SFSpeechRecognizer) and local custom language model functionality to process audio on the teacher’s device, does not send ongoing classroom transcripts through ClassDojo’s LLM Service Providers, and is designed for local command matching rather than transcript summarization — it is the responsibility of each school and district to determine whether and how Live Points may be used consistent with applicable law, including biometric privacy laws, wiretapping and recording-consent laws, FERPA, COPPA, state education laws, and applicable school or district policies. Additionally, Apple’s on-device speech recognition used by both Sidekick Live and Live Points may constitute artificial intelligence (AI) under broad legal definitions adopted by certain states. Schools and districts should consider whether applicable state AI transparency or disclosure laws apply to the use of these features in the classroom. The fact that Live Points avoids the Sidekick Live transcript-to-LLM workflow does not, by itself, resolve those separate legal and policy requirements. The discussion below identifies issues that schools and districts should consider, and the specific Live Points callouts that follow are intended to assist schools and districts in conducting that analysis.

Biometric Data Considerations

This process described above may involve the collection and processing of biometric data as defined under applicable state and federal laws. The audio recording captures vocal characteristics—including pitch, cadence, tone, and other voice features—that may constitute biometric identifiers (such as voiceprints) under various biometric laws and state comprehensive privacy laws that classify voiceprints as biometric data. Although the audio is processed on-device in memory and is never written to an audio file, the initial capture of audio containing these vocal characteristics may constitute the collection or processing of biometric data requiring consent in some jurisdictions.

The legal treatment varies significantly by jurisdiction (e.g. certain states have exclusions to the definition of biometric data for audio or voice recording unless those recordings are used for identification purposes while other states list voiceprints within the definition of biometrics or biometric identifiers and do not have an express exclusion for audio recordings. However, several states in this second group qualify their biometric definitions with language limiting coverage to characteristics “used for the purpose of electronically identifying that person with a high degree of certainty” or similar identification-purpose language. Although these statutes do not contain a standalone audio recording exclusion clause, the identification-purpose qualifier in their definitions may functionally exclude non-identification audio capture such as speech-to-text transcription. Additional states have no current comprehensive privacy law or stand-alone biometrics laws with definitions that specifically address audio data. In these states, federal law (FERPA and COPPA) provides an additional regulatory framework, together with any other applicable state laws and school-specific requirements.

Regardless of state classification, ClassDojo takes the conservative approach of disclosing the potential biometric nature of the audio processing and requiring that the teacher obtain consent in all states where the feature is offered.

Live Points
Live Points, like Sidekick Live, begins with the capture and processing of live voice audio containing vocal characteristics. Although Live Points is not intended to identify a speaker by voiceprint and instead uses speech-to-text plus local pattern matching to detect point-award commands, the initial capture and processing of voice audio may still constitute collection or processing of biometric data in some jurisdictions.

In addition, the student roster names, behavior/skill terms, pronunciations, and other class-specific inputs used to customize local recognition may constitute personal information or student information even though they are used for on-device matching.

The locally generated vocabulary and compiled language model files (specifically, the training data file containing vocabulary and phrase weights, and the compiled speech language model file) may be generated, retained, reused, or cached locally on the teacher’s device during the session and are deleted when the voice command service is torn down (i.e., when the Live Points session ends). These files may contain student roster names, behavior/skill terms, templates, and phrase weights — not audio. No audio files are ever written to disk; all raw audio processing is purely in-memory.

Wiretapping and Recording Consent

Classroom audio recording is subject to applicable state and federal wiretapping and electronic surveillance laws, which impose requirements separate from and independent of biometric consent. Under the federal Electronic Communications Privacy Act (ECPA) (18 U.S.C. § 2510 et seq.), one-party consent is the minimum standard: a party to a communication may record it without the consent of other parties. Because the teacher is present in the classroom and initiates the recording, the teacher likely qualifies as a consenting party under federal law. Classroom recording is also subject to the Family Educational Rights and Privacy Act (FERPA) (20 U.S.C. §1232g), the Children’s Online Privacy Protection Act (15 U.S.C. §§6501–6506) and its implementing Children’s Online Privacy Protection Rule (the “COPPA Rule”) (collectively “COPPA”), and applicable school or district recording policies, each of which imposes independent obligations that must be satisfied regardless of compliance with wiretapping laws.

In addition, states may impose stricter requirements than ECPA. In the 38 one-party consent states, the teacher’s participation in the classroom conversation and affirmative activation of the recording likely satisfies the recording consent requirement under the state wiretapping law. In all-party consent states - such as California, Delaware, Florida, Illinois, Maryland, Massachusetts, Montana, New Hampshire, Pennsylvania, and Washington (as well as Oregon for in-person oral communications and for certain contexts in Colorado, Connecticut, and Nevada) all parties to the recorded conversation must consent where the statute applies to the communication at issue. However, the fact that a recording may be lawful under federal or state wiretapping law does not eliminate other legal and policy requirements. Schools and districts may maintain recording policies that are stricter than state law, even in states where the recording itself is legally permissible.

Teachers are responsible for confirming that their school or district has obtained all necessary recording consents before activating Sidekick Live in an all-party consent state. Wiretapping consent is a separate and independent legal requirement from biometric consent; a teacher may have valid biometric consent but still require separate compliance with applicable wiretapping law.

Live Points
Even if Live Points is designed only to detect point-award commands and to discard unmatched audio locally, the feature still involves live audio capture/listening while active. Accordingly, schools and districts remain responsible for determining whether and under what conditions Live Points may be used consistent with applicable wiretapping and recording-consent laws, notice and consent requirements, school authorization processes, and district policies. The fact that Live Points may avoid cloud transcription or LLM analysis does not, by itself, eliminate those separate legal and policy obligations.

FERPA and Educational Records

FERPA protects student Educational Records from unauthorized disclosure. If a classroom recording captures student voices or any personally identifiable information (PII)—including student names spoken during class, student questions or responses, or other information that is directly linked or linkable to a specific student—the recording and any transcript derived from it may constitute an “education record” under FERPA when maintained by the school or by a party acting for the school. ClassDojo processes the Classroom Audio Transcripts as a “school official” with a “legitimate educational interest” under the school’s or district’s FERPA designation and applicable Student Data Processing Agreement. The school or district may not share or disclose any recording or transcript containing student PII to any party outside of FERPA’s permitted exceptions without the prior written consent of the parent (for K-12 students) or the eligible student. Teachers and school administrators must ensure that the use of this feature, and any resulting Classroom Audio Transcripts, comply with their institution’s FERPA policies. Compliance with state wiretapping law does not satisfy FERPA’s separate requirements.

Live Points
Live Points may process student names spoken aloud, classroom statements, and class roster data. If a resulting point-award event, transcript fragment, or diagnostic log is maintained by the school or by ClassDojo acting for the school and is directly related to an identifiable student, that retained record may constitute an education record under FERPA.

COPPA and Children Under 13

For students under 13 years of age, COPPA requires verifiable parental consent before collecting personal information from children. The COPPA Rule expressly includes (1) audio files containing a child’s voice and (2) biometric identifiers, including voiceprints within the definition of “personal information”. ClassDojo relies on the FTC’s guidance permitting schools to act as the agent of the parent and provide consent on behalf of parents for the collection of student personal information when the collection is solely for educational purposes.

Live Points
For students under 13, Live Points may involve the collection or processing of children’s voice data, audio-derived text, and student roster information for educational purposes. ClassDojo intends Live Points to operate within the same school-authorized educational context described above, but any retained Live Points records should be specifically accounted for in the school’s COPPA analysis and disclosures.

School and District Policies; State Education Code Requirements

Independent of wiretapping laws, school districts may maintain policies requiring written consent from parents or guardians before students may be recorded in the classroom, whether by audio or video. These institutional policies frequently impose requirements that are stricter than applicable state or federal law. Teachers must confirm that their school or district has authorized the use of classroom recording technology and that any required parental notice or consent has been obtained before activating this feature.

The school or district should confirm that it has reviewed and approved the feature under its applicable board policies, vendor contract or student data processing addendum, retention and destruction requirements, and any parent, teacher, principal, or administrator notice or consent procedures required by law or local policy. The school or district also should confirm that no transcript, voice-related data, or student biometric information generated through this feature will be uploaded to, reported to, or maintained in any state system where state law prohibits such reporting or maintenance. Where applicable, this district-level review should also address any state-required local review of privacy implications, civil-rights impact, effectiveness, parental input, and any applicable technology-provider security, return, or destruction obligations.

Teachers should consult their own district’s board policies on recording and technology use before activating Sidekick Live or Live Points, as district policies may impose requirements beyond those of state or federal law. In addition to district-level policies, certain states impose education-code requirements that apply specifically to recording in the classroom. These requirements operate independently of both general wiretapping statutes, biometric consent obligations, FERPA and COPPA.

A teacher’s school or district can disable (opt-out of) the Sidekick Live feature. As part of the district decision process, the authorized district representative should confirm: (a) the district’s Student Data Protection Addendum (DPA) with ClassDojo covers Sidekick Live and Live Points data; (b) compliance with any applicable state recording law, FERPA, COPPA, and school or district policies, including all required parent notices, consents or authorizations under applicable wiretapping, biometric, and education-code laws have been obtained or will be obtained; (c) the district has reviewed the feature’s privacy implications, including any state-specific requirements (such as New York’s local review of biometric technology for privacy implications, civil rights impact, effectiveness, and parental input); and (d) the district’s recording and technology-use policies permit classroom audio transcription for educational purposes.

This district-level authorization step is designed to support that the school or district has evaluated and satisfied all applicable education-code, FERPA, COPPA, and institutional policy, vendor-contract, retention, deletion, and notice/consent requirements before any teacher-level consent is requested or any classroom recording occurs.

Live Points
School and district review should address Live Points separately from Sidekick Live because Live Points raises a different operational question: the device may listen during class for point-award commands, use local speech recognition and class-specific vocabulary, and potentially create retained point-award or diagnostic records even when no general classroom transcript is intended.

As part of district authorization, the school or district should confirm that its contracts, notices, consents, teacher instructions, retention schedule, and deletion procedures expressly cover Live Points, including the use of class roster names, behavior/skill vocabularies, local recognition/model files, silent failures/no-match handling, and the exact data flow when a point is successfully awarded.

What is My Kid in Memories?
My Kid in Memories is an optional feature that uses facial recognition technology through the use of our third party service provider - Amazon (and its Amazon Rekognition product) to help you quickly locate photos of your child in school and classroom photo collections. Instead of scrolling through hundreds of photos, you select a reference photo of your child from your device, and the feature identifies likely matches.

What is a “facial geometry template” and how is it different from a photo?
Think of it like turning a face into a barcode. When you select a reference photo, our third party service provider’s (Amazon) system creates a mathematical representation of your child’s facial features — distances between eyes, nose shape, jawline contour, and other measurements. This “template” is a set of numbers, not a photo. It cannot be reverse-engineered back into an image of your child’s face.

Is this the same as facial recognition used by law enforcement or social media?
No. Law enforcement uses “1-to-many” identification — comparing a face against millions of people. My Kid in Memories uses “1-to-1” comparison — it only checks whether faces in school photos match the specific reference photo you provide. We do not maintain a database of face templates.

What is the difference between “face models or templates” and “face recognition”?
Face models (templates) are like making a barcode for each face — a code that represents the face’s features. Face recognition uses those codes to answer “does this face match that face?” Face detection is different: it asks whether a face is present in a photo and returns face details such as location, landmarks, and pose, but it does not by itself answer whether two faces belong to the same person. An analogy: face models are like sorting socks by pattern (“these two match”), while face recognition is like saying “these socks belong to Dad” because you checked a label. My Kid in Memories creates a face model from your reference photo and then uses recognition to find matching faces in school photos.

What is face detection and how is it different from face recognition?
Face detection analyzes facial geometry and attributes and asks whether a face is present in an image and, if so, where it is and what face details are visible, such as facial landmarks, pose, confidence, and certain facial attributes. We use Amazon’s DetectFaces to determine whether a photo is a group photo.

Face recognition or face comparison asks whether one face matches another face. We use Amazon’s CompareFaces for face comparison when you enable My Kid in Memories. AWS documents DetectFaces and CompareFaces as separate operations, and My Kid in Memories’s matching function uses face comparison rather than face detection alone. For a comparison of DetectFaces and face recognition from Amazon Rekognition – please see here.

Where is my child’s data processed?
The facial geometry template is created using AWS Rekognition’s CompareFaces API. AWS documents this as a “stateless, non-storage” operation — AWS does not persist the face data after processing. If face detection is used, AWS’s DetectFaces API is also documented as a stateless, non-storage operation. ClassDojo has also configured an AI services opt-out policy preventing AWS from using any AI content to develop, improve, or train its models. The reference photo is not separately stored by ClassDojo for this feature, but it may be temporarily transmitted to AWS Rekognition to perform the comparison you request.

Does ClassDojo store my child’s face template?
No. Templates are retained only transiently — for seconds — while the comparison is computed (e.g. for the active matching session). Once match results are returned, templates are deleted. ClassDojo does not maintain a permanent database of children’s face templates.

What about the other children in the school photos?
The system briefly creates temporary “exclusion templates” for other faces detected in the searched photos solely to exclude those faces from being returned as matches. These are deleted within seconds and never stored, shared, or used for any other purpose. AWS may also use DetectFaces to determine whether faces are present in the searched images and to return face details needed for that processing. DetectFaces itself does not perform face matching.

How can I opt-out if I do not want an exclusion template created of my child?
If your child appears in school or classroom photos and you do not want ClassDojo (through our third-party service provider - Amazon) to create exclusion templates from your child’s face for future My Kid in Memories searches, you may submit a suppression request at privacy@classdojo.com. Once we process the request, we will suppress your child’s applicable school, classroom, or photo set so that new exclusion templates are not created for your child for those future searches, except to the limited extent technically necessary to maintain and honor the suppression request. Because exclusion templates are transient and deleted within seconds, this opt-out operates prospectively and does not reverse processing that occurred before the suppression request was received.

If your child appears in school or classroom photos made available through ClassDojo, ClassDojo (through its service provider, AWS using the Amazon Rekognition product) may create temporary mathematical representations of your child’s facial features from those photos solely to exclude your child from being returned as a match when another parent uses the My Kid in Memories feature. These temporary “exclusion templates” are not used to identify or match your child across photos, are not used to enroll your child in My Kid in Memories, and are deleted within seconds after the exclusion comparison is complete. ClassDojo does not maintain a persistent database of exclusion templates for this purpose. AWS’s CompareFaces API is a stateless, non-storage operation, and ClassDojo has configured an AI services opt-out policy with AWS instructing AWS not to use or store AI content processed through Rekognition to develop, improve or train that service or technologies of AWS or its affiliates.

Will my child’s biometric data be used to train AI?
No. ClassDojo has implemented an AWS AI services opt-out policy prohibiting AWS from using any AI content processed through Rekognition for model training or service improvement. ClassDojo will never use your child’s biometric data to train any of ClassDojo’s AI systems.

Does ClassDojo sell my child’s biometric data?
No. ClassDojo does not sell, lease, trade, or otherwise profit from your child’s biometric data. The only entity that processes the data is AWS, solely to perform the facial comparison under a written agreement that strictly limits its use.

What if I don’t want to use this feature?
My Kid in Memories is completely optional and off by default. You can use Browse All Photos to manually view every school photo, create photo books, and place orders without enabling facial recognition.

How do I withdraw my consent?
You can withdraw at any time (Settings > Privacy > My Kid in Memories > Withdraw Consent) or by emailing our Privacy Team at privacy@classdojo.com. My Kid in Memories photo matches will be disabled immediately. Because templates are only retained during active matching, there are usually no active templates remaining at the time of withdrawal. Any residual biometric-related metadata, queued artifacts, and local cached references (if any) under our control will be permanently destroyed within 15 days of the request.

Is there a separate consent for COPPA?
Yes. ClassDojo obtains a separate COPPA-compliant parental consent in addition to this biometric consent. The two consents address different legal requirements but work together. The COPPA consent also links to this dedicated Biometric Privacy page.

Can I get a copy of my consent record?
You can view or download a PDF copy here.

What is Sidekick Live?
Sidekick Live is an optional feature that captures classroom audio (including student voices) on the teacher’s device, converts it to a text transcript using Apple’s on-device speech recognition engine, implemented using Apple’s SpeechAnalyzer/SpeechTranscriber API and transmits the text transcripts to ClassDojo for further AI-powered analysis with ClassDojo’s LLM Service Providers to provide teaching insights, discussion summaries, and educational observations. ClassDojo does not collect or process any audio recordings.

Why is this considered biometric data?
Audio recordings capture vocal characteristics—pitch, cadence, tone, and other voice features—that may be classified as biometric identifiers (voiceprints) under certain state laws ClassDojo requires that schools obtain consent in all states where the feature is offered.

Does the classroom audio recording leave my device?
No. The audio is processed entirely in memory on your device using Apple’s on-device speech recognition engine and is not sent to ClassDojo, Apple, or any third party as an audio file. Only the resulting text transcription is transmitted to ClassDojo. Because no raw audio recording file is created, there is no audio file for the teacher to delete after the session.

Does Apple train on my audio or transcript?
No. The audio is processed entirely on your device using Apple’s Speech framework (SpeechAnalyzer / SpeechTranscriber) and is processed solely in memory - no audio file is created. Only the resulting text transcription is transmitted to ClassDojo. Apple’s published SpeechAnalyzer materials state that this transcription path uses an on-device model, although the relevant model assets may be downloaded to and retained in system storage on the device.

What is Zero Data Retention (ZDR)?
ZDR is a contractual and technical commitment from our cloud LLM Service Providers.

ClassDojo has entered into a specific contract with the LLM Service Providers for “zero data retention” beyond the transient retention for the ClassDojo-provided AI tools mentioned above. This means that the prompts, inputs (e.g. the Classroom Audio Transcripts), and outputs (e.g. the summaries of the Classroom Audio Transcripts) (1) will not be logged for human review by the LLM Service Providers; and (2) will not be retained by the LLM Service Provider. Note, however, that the LLM Service Providers may perform automated screening of the inputs and prompts for safety purposes (the “Safety Classifiers”), which shall consist solely of metadata (including classifier types, dates, counts, and confidence scores) and shall not include any inputs, prompts, or outputs themselves (including summarizations of prompts, inputs, or outputs) or any portion thereof. LLM Service Providers may retain data from Safety Classifiers.

Does the Apple Speech to Text framework or the ClassDojo LLM Service Providers identify which student said what?
No. The transcription captures classroom dialogue as undifferentiated text. ClassDojo does not use voice recognition or voiceprint matching to attribute statements to identified students. However, student names or identifiers may appear in the transcript if they are spoken aloud during class (e.g., “Good answer, Sarah”). Teachers should be aware that the transcript may contain incidentally captured student names.

Are student voiceprints created?
No. Apple’s on device speech framework (SpeechAnalyzer / SpeechTranscriber) engine performs speech-to-text conversion—it does not create voiceprint templates, speaker identification models, or voice signatures. The system converts audio content to text; it does not map voice characteristics to individual identities. To ClassDojo’s knowledge, Apple’s on-device speech recognition engine does not create persistent voiceprint templates or speaker identification models as part of the speech-to-text conversion process.

What about students in all-party consent states?
In states that require all-party consent for recording (e.g. California, Delaware, Florida, Illinois, Maryland, Massachusetts, Montana, New Hampshire, Pennsylvania, Washington, and Oregon for in person oral communications), the school or district must authorize classroom transcription and provide notice to parents or guardians. In California, school principal consent is independently required under Education Code §51512. Similar laws may apply in other states. The teacher consent flow includes a state-specific acknowledgment for all-party consent states.

Does ClassDojo store or sell voice data or transcripts?
No. ClassDojo does not store, sell, lease, trade, or otherwise profit from voice data, audio recordings, or Classroom Audio Transcripts. ClassDojo does not store raw classroom audio at all. Raw audio is processed on the teacher’s device and is not written to an audio file. Classroom Audio Transcripts are retained only transiently by ClassDojo so that they can be transmitted to, and processed by, our LLM Service Providers. The only third party that processes the Classroom Audio Transcript is these LLM Service Providers under a ZDR Agreement with processing occurring under strict contractual limitations prohibiting any secondary use. ClassDojo may store the summaries of the Classroom Audio Transcripts (e.g., the output produced by the LLM Service Providers), subject to the applicable school data protection agreement and deletion process.

Is there a separate consent for COPPA?
ClassDojo relies on the COPPA school consent exception, which allows schools to authorize operators to collect students’ personal information for educational purposes in lieu of direct parental consent. The teacher’s biometric consent (for audio capture) or wiretapping consent is separate from and in addition to the COPPA school consent for transcript data.

Can I see the sessions I authorized?
Yes. A copy is sent to your email each time you provide authorization for a new session. You can also view or download a copy of your consent history at any time (Settings > Privacy > Sidekick Live > View Consent History).

What is Live Points?
Live Points is a separate optional classroom feature intended to let a teacher award points or trigger behavior acknowledgments through spoken commands while class is in session. It is designed around on-device speech recognition plus local pattern matching rather than transcript summarization.

How is Live Points different from Sidekick Live?
Sidekick Live is designed to create a text transcript of classroom discussion and send that transcript to ClassDojo and then to ClassDojo’s LLM Service Providers for further analysis. Live Points is intended instead to listen for point-award commands, use local speech recognition and local matching logic, and avoid the Sidekick Live transcript-to-LLM workflow.

Does Live Points use LLMs or generative AI?
No. Live Points does not use generative AI, LLMs, or ClassDojo’s LLM Service Providers to analyze classroom audio. Live Points does use Apple’s on-device speech recognition, which may constitute AI under broad legal definitions, combined with local rule-based pattern matching.

What does “pattern matching” mean?
Pattern matching means the software compares locally generated words or phrases against expected student names, behavior/skill terms, aliases, pronunciations, and command templates and then applies simple rules or confidence thresholds to select the most likely intended match. It is not generative AI and it is not voiceprint identification.

What information is used to improve matching accuracy?
Live Points may use class-specific inputs such as student roster names, behavior/skill lists, aliases, pronunciation data, and command templates to improve recognition accuracy for that class.

Does Live Points use a wake word?
Live Points does not use a wake-word detector. While the feature is active, Apple’s on-device speech recognition continuously converts live classroom audio into recognized words and phrases on the teacher’s device, and the ClassDojo app locally checks those recognized words and phrases for a matching command pattern. In other words, Live Points uses continuous on-device speech recognition plus local command matching; it does not listen only for a single trigger phrase and ignores everything else.

Does the audio or transcript leave my device?
Raw Live Points audio remains on the device during local speech recognition and custom language model processing. If a valid point-award command is detected, ClassDojo receives only the same structured point-award data that would be sent during a manual point award. No transcript text, phrase snippets, confidence scores, or speech metadata are sent to ClassDojo servers as part of that point-award request. Live Points does not send an ongoing classroom transcript through the Sidekick Live transcript-to-LLM workflow.

Are student voiceprints created?
Live Points is not intended to create speaker-identification templates or voiceprint profiles. It is intended to convert speech into local text and compare that text to expected command patterns. However, the capture and processing of voice audio may still be treated as biometric data under some laws.

What happens if the system cannot match a student or behavior?
If Live Points cannot confidently identify the intended student (or group) and the applicable behavior or skill, it does not award points. Instead, the current behavior is a silent failure or “no match” result.

What happens when the system does find a match?
When the system finds a valid match, it generates the same structured point-award event data that would be sent if the teacher had manually tapped to award the point, such as the relevant student or group identifier, point value, and behavior category. No transcript text, phrase snippets, confidence scores, or speech metadata are sent to ClassDojo servers as part of that point-award request.

Are roster names and behavior lists stored on the device?
Live Points may create temporary on-device vocabulary and speech-model files to improve recognition accuracy. These files help the device recognize classroom-specific names and commands more accurately and may be retained, reused, or cached locally on the device during the session. They may contain roster names, behavior terms, pronunciations, templates, and related language-model data — not audio — and are deleted when the voice command service is torn down at the end of the session. No audio files are ever written to disk.

How do I delete Live Points data?
Because Live Points relies primarily on local processing, the main retained records, if any, are the resulting point-award records and related logs. Schools and districts may request deletion of retained Live Points-related records by contacting privacy@classdojo.com or through the applicable Student Data Protection Addendum process.

Is there a separate consent for COPPA?
ClassDojo relies on the COPPA school consent exception, which allows schools to authorize operators to collect students’ personal information for educational purposes in lieu of direct parental consent. The teacher’s biometric consent (for audio capture) or wiretapping consent is separate from and in addition to the COPPA school consent for transcript data.